London developers contribute community detection code for GraphLab

More good news: we just got additional code contribution, this time from Michael Leznik and Ryan Toppings. They have implemented the following Community detection algorithm based on label propagation. LPA-algorithm is explained in http://arxiv.org/pdf/0910.1154.pdf :

Advanced modularity-specialized label propagation algorithm for detecting communities in networks. X. Liu, T. Murata Tokyo Institute of Technology, 2-12-1 Ookayama, Meguro, Tokyo 152-8552, Japan

The contribution contains Java code for GraphChi Java version. (One can find also C++ code for GraphChi C version )

The required Java files are:

BidirectionalLabel.java

CommunityAnalysis.java

CommunityDetection.java

graph_test_data.tsv

graph_test_data.tsv.communities_ids

The code contains a toy example, which one might run passing the following parameters to CommunityDetection class which implements GraphChiProgram<Integer, Integer>

graph_test_data.tsv 2 edgelist

The first parameter is graph file name

The second parameter is number of shards used

The third parameter is the type of graph used

Don't forget to link GraphChi to the project 

It outputs the file in this case: graph_test_data.tsv.communities_ids which contains labels id and label ids belonging to this label. Output for this is included in the zip file.

Thanks a lot Michael & Ryan - you contributions are highly appreciated!

Kobo contributes open source LDA inference code for GraphLab

More good news to report! Just heard from our collaborators at Kobo: Curtis Ge and Darius Braziunas that their code contribution is cleared. Kobo released a Latent Dirichlet Allocation (LDA) code which computes inference based on GraphLab LDA model.

In more detail, GraphLab LDA trains the model and classifies existing documents into clusters but when new documents are obtained there is no way for assigning them to the learned classes. Kobo's code classifies new documents based on GraphLab LDA model.

Anyone who wants to test Kobo's code is welcome to try it out. We will soon merge this code into the main GraphLab branch.

Interesting Twitter Graph Analysis & Visualization

I got this from my colleage Zach Nation: an interesting Twitter graph visualization.

Cisco mentions that they are using GraphLab as part of their security stack

Read more here.

O'Reilly Blog Post on GraphLab SFrames

Fresh out of the oven: Ben Lorica from O'Reilly just published a blog post which gives visibility to our new functionality: SFrames.
Additionally this blog post also appears at Forbes.

Drug repurposing using GraphLab

Recently I learned about an interesting work by Murat Can Cobangolu, a graduate student at the CMU-Pitt computational biology institution who is using GraphLab for predicting drug target interaction.
The basic idea is that a new medication creation is a complicated and costly process. The idea is that we can use drugs that are already safe to use for other purposes relative to the ones they were designed. For example, Viagra was design to be heart medication and eventually ended up as a stimulator..

The work is reported in the following paper: 

• MC Cobanoglu et al., (2013) "Predicting Drug-Target Interactions Using Probabilistic Matrix Factorization." Journal of Chemical Information and Modeling, vol. 53, pp. 3399–3409. http://pubs.acs.org/doi/ipdf/10.1021/ci400219z

The construction is rather simple. A matrix of drugs vs. their gene interaction is construction and a matrix factorization methods is built to build a model. Using the model one can predict the interaction of any drug with any gene.

If you like to learn more about this work, and GraphLab applications to Health Care, you are invited to attend our 3rd GraphLab Conference. We have invited Murat to present his work there.

Spotlight: Treato - big data analytics for patient reaction to medication

I stumbled upon this interesting Israeli startup company who is analyzing patient reaction to medications.They have a large dataset of crawled patient text comments about the drugs they are using and the apply some fancy analytics. For example, you can search for a specific medication like Tylanol

Quickly get an overview about it:

Learn how user rank this medication: 

Additionally you can learn about people concerns:

Very cool application utilizing very big data!